LIMNOLOGY

and Limnol. Oceanogr. 63, 2018, S99–S118 VC 2017 The Authors Limnology and Oceanography published by Wiley Periodicals, Inc. OCEANOGRAPHY on behalf of Association for the Sciences of Limnology and Oceanography doi: 10.1002/lno.10726

The metabolic regimes of flowing waters

E. S. Bernhardt ,1* J. B. Heffernan ,2 N. B. Grimm ,3 E. H. Stanley ,4 J. W. Harvey ,5 M. Arroita,6,7 A. P. Appling,8 M. J. Cohen,9 W. H. McDowell ,10 R. O. Hall, Jr.,7,a J. S. Read,11 B. J. Roberts,12 E. G. Stets,13 C. B. Yackulic14 1Department of Biology, Duke University, Durham, North Carolina 2Nicholas School of the Environment, Duke University, Durham, North Carolina 3Arizona State University, Tempe, Arizona 4Center for Limnology, University of Wisconsin, Madison, Wisconsin 5National Research Program, U.S. Geological Survey, Reston, Virginia 6Department of Plant Biology and Ecology, University of the Basque Country, Bilbao, Spain 7Department of Zoology and Physiology, University of Wyoming, Laramie, Wyoming 8US Geological Survey, Office of Water Information, Tucson, Arizona 9School of Forest Resources and Conservation, University of Florida, Gainesville, Florida 10Department of Natural Resources and the Environment, University of New Hampshire, Durham, New Hampshire 11U.S. Geological Survey, Office of Water Information, Middleton, Wisconsin 12Louisiana Universities Marine Consortium, Chauvin, Louisiana 13U.S. Geological Survey, National Research Program, Boulder, Colorado 14U.S. Geological Survey, Southwest Biological Science Center, Flagstaff, Arizona

Abstract The processes and biomass that characterize any ecosystem are fundamentally constrained by the total amount of energy that is either fixed within or delivered across its boundaries. Ultimately, ecosystems may be understood and classified by their rates of total and net productivity and by the seasonal patterns of pho- tosynthesis and respiration. Such understanding is well developed for terrestrial and lentic ecosystems but our understanding of ecosystem phenology has lagged well behind for rivers. The proliferation of reliable and inexpensive sensors for monitoring dissolved oxygen and carbon dioxide is underpinning a revolution in our understanding of the ecosystem energetics of rivers. Here, we synthesize our current understanding of the drivers and constraints on river metabolism, and set out a research agenda aimed at characterizing, classi- fying and modeling the current and future metabolic regimes of flowing waters.

The fuel that powers almost all of Earth’s ecosystems is autotrophs and heterotrophs) is measured as ecosystem res- created by organisms capable of the alchemy of photosyn- piration (ER). Together, GPP and ER are the fundamental thesis, in which solar energy, water, and carbon dioxide are metabolic rates of ecosystems that constrain the energy sup- converted into reduced carbon compounds that are then ply and energy dissipation through food chains, and the bal- used to sustain life. We measure this conversion of solar ance of these two fluxes, measured as net ecosystem energy into organic energy as the gross primary productivity production (NEP), determines whether carbon accumulates (GPP) of ecosystems. The collective dissipation of this or is depleted within an ecosystem. Terrestrial ecosystems organic energy through organismal metabolism (of both often have predictable annual cycles, with both GPP and NEP typically peaking during warmer and wetter months of the year. In many well-studied lakes productivity peaks *Correspondence: [email protected] when warming temperatures, lengthening days, and high aPresent address: Flathead Lake Biological Station, University of Montana, nutrient concentrations occur in concert. The life cycles of Polson, Montana many consumers are likely synchronized to these seasonal oscillations such that periods of peak energetic demand by This is an open access article under the terms of the Creative Commons consumers coincide with or follow the peak productivity of Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly their preferred plant or prey (e.g., Lampert et al. 1986; Berger cited and is not used for commercial purposes. et al. 2010). As a result, ecosystem respiration tends to

S99 Bernhardt et al. Metabolic regimes

Fig. 1. Temporal trends in dissolved oxygen (DO shown as % of atmospheric saturation) for four contrasting U.S. rivers over a 4-yr period. High diel variation in dissolved oxygen is a proxy for high ecosystem GPP. The top trace is from the Menominee River in northern, Wisconsin, a large river with clear summer peaks and winter lows in stream metabolism. Dissolved oxygen traces for the more southern Five Mile Creek in Alabama and San Anoto- nio River in Texas show little seasonality, with the smaller Creek having sustained high diel variation in DO and the more urban river having many alternating periods of high and low diel DO variation. The bottom trace is from Fanno Creek, a heavily shaded and frequently flooded small stream in western Oregon. Data from each of these four streams appears again (along with axes labels) in Figs. 3, 5. Here, we remove axes scores to focus on the differences in seasonality of a common signal across streams. The scale of both axes is the same for all four series. covary with GPP. This phenology of the ecosystem, or its reduced coherence between climate drivers and river ecosys- seasonal timing of carbon, water and energy exchange (sensu tem productivity and respiration. Noormets 2009) is both a cause and a consequence of the Of course, the relative importance of terrestrial shading phenology of all component organisms. and terrestrial organic matter inputs varies with river size The same “cause and consequence of organism (Vannote et al. 1980). Just as the importance of terrestrial phenology” argument cannot be made for river ecosystems inputs of nutrients and organic matter to lake food webs for three reasons. First, in many rivers seasonal variation in diminishes as the ratio of lake size to watershed size light is uncorrelated with seasonal variation in temperature, increases (Tanentzap et al. 2017), we expect that the relative because of the reduction in light supply due to canopy inter- importance of canopy shading, allochthonous inputs, and ception, sediment loads or colored organic matter. Second, hydrologic disturbance should decline between headwater in many rivers intense and frequent high flow events regu- streams and large rivers. This gradient in river size impacts larly reduce the biomass of autotrophs (algae, mosses, and the expected ecosystem phenology, with well-lit and less fre- macrophytes) through scouring or burial while stream drying quently disturbed large rivers having regular summer pro- can strand and desiccate autotrophs on the channel bed. ductivity peaks while shaded headwaters with frequent Finally, most rivers receive energetic subsidies in the form of flooding are likely to have productivity peaks that are mis- detritus and dissolved organic matter from their surrounding matched to the terrestrial growing season (Fig. 1). These pre- watersheds. These allochthonous inputs can match or exceed dicted longitudinal patterns can be obscured for rivers with in situ GPP and thus decouple the seasonal and annual pat- high sediment or colored organic matter inputs. Because the terns of GPP and ER. For each of these reasons we expect a temporal signals of GPP and ER are so diverse across streams

S100 Bernhardt et al. Metabolic regimes and years, and so often asynchronous with terrestrial pro- in the concentration of dissolved oxygen (DO) throughout a ductivity and climate drivers, we propose that the term meta- diel (24-h) cycle, using increases during daylight hours and bolic regimes is more appropriate than phenology to describe overnight declines to calculate rates of GPP, ER, and NEP. In these patterns. Here, we define a metabolic regime as the Odum’s initial metabolism estimates (Odum 1956, 1957) characteristic temporal pattern of ecosystem GPP and ER these changes were documented by collecting samples at 2– observed for a river. 4 h intervals throughout a day followed by manual analysis Despite the frequent mismatches in the timing of peak of DO concentration via titration. Sampling around the energy supply, thermal optima, and disturbance river ecosys- clock by this method is labor intensive, and the duration of tems support a tremendous diversity of species (Strayer and early studies was thus limited to a handful of days during Dudgeon 2010) and can convert enormous quantities of the year. Later efforts were enabled by instantaneous meas- organic matter and inorganic nutrients into CO2,CH4,N2, urements of DO with first generation environmental sensors. and N2O gases (Cole et al. 2007; Mulholland et al. 2008; Bat- Because these sensors were expensive and required frequent tin et al. 2009; Raymond et al. 2013; Stanley et al. 2016). calibration, researchers tended to deploy them for very lim- The capacity of rivers to support these critical functions ited periods of time. As a result, most reported rates of river depends, fundamentally, on ecosystem metabolism, defined ecosystem metabolism were derived from a small number of as “the production and destruction of organic matter, and the measurement days over a year (e.g., 2–12 d; synthesized by associated fluxes of nutrients, through the gross photosynthetic Lamberti and Steinman 1997; Finlay 2011). These brief sam- and respiratory activity of organisms” (Odum 1956). Although pling regimes are biased toward optimum field conditions H. T. Odum developed the general concept and made the for sensor deployment (e.g., sunny days with low flows). first measurements of whole ecosystem metabolism in a river They are thus unlikely to provide accurate estimates of mean over 60 yr ago (New Hope Creek, North Carolina, Odum or annual metabolic rates, and are almost certainly insuffi- 1956), stream ecologists have made remarkably little progress cient for capturing the impacts of hydrologic disturbance or in building our empirical understanding of stream metabo- pulsed resource supply. lism (McDowell 2015). Fortunately, the proliferation of more rugged and less It is perhaps not surprising, given the extremely dynamic expensive environmental sensors is now allowing river scien- nature of river metabolism, that we have uncovered no uni- tists to overcome the traditional logistical challenges of mea- versal or even regional predictors of river productivity. suring metabolism. Looking forward, we should now be able Unlike terrestrial ecosystems, where rainfall and temperature to address fundamental questions about how river ecosys- explain much of the global variation in primary production tems function, such as: What controls the variation in (Whittaker 1962; Leith 1975; Field et al. 1998), or lentic eco- the magnitude and timing of productivity within and systems, in which nutrient loading and organic matter are among rivers? How are these controls changing in primary drivers of productivity (Schindler 1978; Lewis et al. response to climate or land use change? How will 2011), attempts to uncover these same patterns in streams resulting changes in river ecosystem productivity con- have proven elusive (Lamberti and Steinman 1997; Bernot strain their capacity to support freshwater biodiver- et al. 2010). Ironically, there is a rich history of conceptual sity, food production, and the maintenance of water models predicting spatial patterns of metabolism in streams quality? In this paper, we hope to help shape and catalyze and rivers beginning with the River Continuum Concept this research frontier by summarizing our current under- (RCC; Fig. 2) (Vannote et al. 1980). Tests of this and other standing of river metabolism; proposing a series of new and models have been limited and often equivocal (e.g., Bott fundamental research questions; and providing a series of et al. 1985; Minshall et al. 1992; McTammany et al. 2003). A examples demonstrating exciting new applications of contin- literature synthesis showed that watershed area (and thus uous metabolism datasets. river size) predicted the magnitude of annual GPP in rela- tively undisturbed watersheds, but had little predictive What do we know so far? power for GPP in even moderately developed watersheds, or for ER across all rivers (Finlay 2011). Other syntheses at daily How we measure and model river metabolism rather than annual timescales have found even less predict- Stream ecosystem metabolism (in units of oxygen, g O2 22 21 ability in metabolic rates (Mulholland et al. 2001; Bernot m d ) is estimated from the daily variation in the produc- et al. 2010; Hoellein et al. 2013). tion and consumption of oxygen by stream organisms fol- We attribute our limited success in uncovering patterns or lowing a simple model: building predictive models of river ecosystem metabolism to dDO GPP1ER 5 1KðÞDO 2DO the challenging combination of technological constraints dt z sat and the dynamic physical environment characteristic of dDO many rivers. In theory, the measurement of river metabolism where dt is the change in dissolved oxygen concentration is straightforward. An investigator simply measures changes through time, GPP is the rate of photosynthetic O2 production,

S101 Bernhardt et al. Metabolic regimes

Fig. 2. A conceptual model depicting differences in how climate, light and hydrologic regimes vary along the river continuum and between three terrestrial biomes. The climate diagrams across the top show average monthly precipitation in blue bars with daily air temperatures shown as blue (minimum) and red (maximum) lines.

ER is the rate of oxygen consumption through both autotro- (Fig. 3A), except when high rates of gas exchange or flow varia- phic and heterotrophic respiration and KðÞDOsat2DO is the net tion obscure this pattern. Variation in the shape of this oxygen exchange of oxygen between the water column and the overly- curve reflects variation in rates of GPP, ER, and K thus, with suf- ing air and is governed by a per unit time gas exchange rate K. ficient rates of GPP, it is possible to use curve fitting routines Mean river depth (z) converts from volumetric to areal rates. By to estimate all three parameters from the oxygen data convention, any process that lowers DO concentration in the (Holtgrieve et al. 2010; Hall 2016) available for each day within water takes on a negative value so that ER is always a negative a time series. number. Whenever autotrophs are present, GPP will increase A first wave of sensor-enabled studies using this model- with light. Typically, GPP results in peak DO during the day ing approach clearly documented that light availability and

S102 Bernhardt et al. Metabolic regimes

Fig. 3. Diel and Storm Recovery Patterns in Rivers. Figures in row A are conceptual models representing the typical variation in O2 (as % saturation) at diel (A1, A2) and over storm-recovery trajectories (A3). Figures in row B show three contrasting diel curves selected from each of the four rivers shown in Fig. 1. In row C, we show 60 d of estimated rates of gross primary productivity (in green, derived from the O2 data in Fig. 1) alongside the river hydrograph (gray shading). Each 60-d period encompasses at least one major flood. The color and date of the triangles shown at the top of pan- els in row C correspond to the dates for the diel O2 curves shown in row B.

flooding are dominant and countervailing drivers of river use change is having less of a direct effect on river metabo- ecosystem productivity (Uehlinger and Naegeli 1998; lism than are the often-accompanying reductions in ripar- Houser et al. 2005; Roberts et al. 2007; Izagirre et al. 2008; ian canopy shading and increases in the frequency and Beaulieu et al. 2013; Hope et al. 2014; Huryn et al. 2014; severity of floods or droughts. Indeed, these alterations to Roley et al. 2014). In contrast to lakes there is thus far lim- river light and flow disturbance regimes are likely changing ited evidence that nutrient availability controls metabolic the potential for rivers to perform the critical ecosystem ser- rates in rivers (Hoellein et al. 2013; Solomon et al. 2013). In vice of nutrient assimilation and retention (Fellows et al. contrast, variation in stream metabolic rates can affect 2006; Mulholland et al. 2008). Those alterations that reduce nutrient dynamics both within and across rivers (Hall and annual GPP and ER or which shift the timing of peak activ- Tank 2003; Roberts and Mulholland 2007; Heffernan and ity away from peak nutrient loading are likely to exacerbate Cohen 2010; Lupon et al. 2016). It thus seems likely that nutrient pollution by reducing the potential for instream widespread nutrient pollution of rivers resulting from land processing.

S103 Bernhardt et al. Metabolic regimes

The light and thermal regimes of rivers river ecosystems relative to their terrestrial or lentic counter- For terrestrial ecosystems, light and thermal regimes are parts (e.g., Huryn et al. 2014). Ecosystem respiration within controlled by latitude, topography and precipitation pat- rivers appears to have stronger temperature sensitivity than terns, and thus two very basic descriptors of climate, mean GPP, leading some to suggest that rivers are poised to annual temperature and mean annual precipitation, can become increasingly large sources of CO2 to the atmosphere explain much of the variation in GPP across terrestrial bio- in a warming climate (Demars et al. 2011a,b). mes (Whittaker 1962). Unfortunately, the light and thermal Because terrestrial vegetation both reduces light and regimes of streams are less easily linked to these widely avail- enhances organic matter loading to rivers, we are unlikely to able environmental data. The light that penetrates from the see strong coherence between the productivity of rivers and atmosphere to river surfaces is strongly affected by both the productivity of their surrounding terrestrial biome. channel orientation with respect to landscape features (e.g., Indeed, regions of high terrestrial primary productivity (e.g., canyon walls or banks) and shading by terrestrial vegetation tropical rain forests) are often associated with very low light (Vannote et al. 1980; Hill and Dimick 2002; Julian et al. availability for small receiving streams (Mulholland et al. 2008). For small channels, the phenology of terrestrial vege- 2008), while low productivity terrestrial ecosystems (e.g., tation can lead to rapid transitions in the light regime dur- deserts, tundra) can be drained by some of the most produc- ing the leaf out and litterfall periods that are particularly tive streams in the world (Grimm 1987). Terrestrial vegeta- pronounced in temperate streams (Hill and Dimick 2002). tion exerts less influence on the light regimes of larger rivers Once light reaches the river surface there is a predictable in which the attenuation of light by water depth and turbid- attenuation with water depth that can be exacerbated by the ity become dominant constraints (Vannote et al. 1980). Sim- reflectivity of suspended materials and light absorption by ilarly, as rivers increase in size, their thermal regimes dissolved organic matter (Davies-Colley and Smith 2001; become increasingly insensitive to the stature and status of Julian et al. 2008). Light attenuates exponentially with water surrounding terrestrial vegetation. Thus, the position of any depth, such that benthic light availability is highest during river segment within its river network is likely to be as baseflow periods and lowest during high flows. Moreover, important a determinant of its metabolic regime as the ter- light attenuation tends to increase in the downstream direc- restrial biome which it drains. The distance along the river tion as larger rivers are deeper than their headwaters (Julian continuum at which terrestrial influence wanes is likely to et al. 2008). For all these reasons, light availability varies sub- vary considerably among different terrestrial biomes (Fig. 2). stantially among sites as well as over time, well beyond the For example, riparian canopy cover is typically absent from latitudinal variation in solar radiation that is the primary the headwaters of desert, grassland, and alpine streams and driver of variation in light regime among terrestrial ecosys- seasonal light regimes are considerably different for small tems. Because of these light filtering mechanisms, periods of streams draining temperate deciduous vs. tropical or ever- high light availability may not be matched to periods of green forests despite their similar channel widths. These high temperature or high incident solar radiation. Indeed, interactions between geomorphology, vegetation, and chan- for many small, forested streams of the temperate zone, the nel size will drive wide variation in both the annual magni- highest light availability occurs in early spring before the for- tude and the timing of primary production and ecosystem est canopy leafs out and late autumn after litterfall (Roberts respiration. et al. 2007). Like the light regime, the thermal regime of rivers can Disturbance regimes in river ecosystems deviate substantially from the temperature of the overlying Though much of what we have learned about annual pat- atmosphere. The temperature of a river at any given time is terns of river metabolism has come from studies of spring determined by the cumulative fluxes of energy into and out fed or regulated streams (Odum 1957; Roberts et al. 2007; of the water, including groundwater, evaporation, and sun- Heffernan and Cohen 2010), a minority of rivers have such light (Poole and Berman 2001). In general, river tempera- predictable flows. The potential for rivers to support high tures will fluctuate seasonally in concert with air rates of GPP is maximized in slow flowing, clear water temperature, but the average temperature and the magnitude streams with stable flow and high light availability where of diel and seasonal swings in temperature will be consider- productivity can rival that of temperate forests (Odum 1957; ably muted in rivers relative to the overlying atmosphere. Heffernan and Cohen 2010; Acuna~ et al. 2011). In contrast, For rivers with significant rates of water and heat exchange rivers with frequent bed-moving events typically have signif- with groundwater, or rivers below dams that receive most of icantly lower productivity, even under similarly high light their flow from reservoir depths there can be little to no cor- regimes (Uehlinger and Naegeli 1998; Uehlinger 2006). A relation between air and water temperatures over time (Poole rapidly growing body of literature documents the overriding and Berman 2001; Olden and Naiman 2009). The resulting importance of high-flow or desiccation-initiated disturbances mismatches between temperature and light generate a much of benthic primary producers in structuring the magnitude more diverse combination of light and thermal regimes in and timing of river ecosystem productivity (Acuna~ et al.

S104 Bernhardt et al. Metabolic regimes

2004, 2005; Atkinson et al. 2008; O’Connor et al. 2012; et al. 1997) or activate immediately in others (Antarctic Beaulieu et al. 2013). streams, McKnight et al. 2007). There has been limited study Floods can regulate spatial and temporal patterns of many to date of the ecosystem respiration rates of dry riverbeds, ecological processes in rivers including metabolism (Fig. 3). but one might expect enhanced mineralization of buried The return interval for events that lead to reductions in liv- organic matter as anoxic areas of the sediments are oxygen- ing biomass or organic matter stocks in rivers is much ated during drydown as well as pulses of microbial activity shorter than for almost any other ecosystem type (Grimm following rewetting of desiccated sediments (Merbt et al. et al. 2003). In the most extreme cases forests may be har- 2016), analogous to the pulse of mineralization often vested on decadal time scales and grasslands may burn annu- observed when rain falls on dry soils (Fierer and Schimel ally, but some rivers experience dozens of biomass-reducing 2002). disturbances per year. A large flood can have multiple effects Nutrient impacts on stream metabolic regimes on primary producers and the macro- and microconsumers When light and disturbance are not limiting, nutrient that live on or within riverbed sediments. Floods that are supply may limit the magnitude or influence the phenology sufficiently powerful to mobilize bed sediments may lead to of river metabolism (Hill et al. 2009). Common anthropo- burial, scour or export of benthic producers and organic mat- genic impacts on rivers such as the removal of riparian vege- ter (Grimm and Fisher 1989; Young and Huryn 1996; Ueh- tation and flow regulation should exacerbate the sensitivity linger and Naegeli 1998; Biggs et al. 1999; Uehlinger 2000; of metabolism to nutrient loading by reducing light limita- Uehlinger et al. 2002; Atkinson et al. 2008). It is literally true tion and disturbance frequency as constraints on river eco- that “the rolling stone gathers no moss1,” as mobile substrates system productivity. There are plentiful examples of nutrient are unable to accumulate large and stable amounts of ben- limitation within individual rivers (Grimm and Fisher 1986; thic biomass. Even in the absence of physical disturbance of Francoeur 2001; Tank and Dodds 2003). Experiments to channel sediments, the deeper and more turbid waters typi- measure the impact of nutrient loading on river metabolism cal of floods decrease light availability to the channel bed have also documented substantial increases in GPP for a river (Hall et al. 2015). Research to date in small streams suggests in Alaska fertilized with phosphorus (Peterson et al. 1985) that recovery times for both GPP and ER are expected to be and enhanced rates of ER under N 1 P enrichment in heavily longer for large floods that exceed the critical threshold for shaded, low nutrient streams in North Carolina (Kominoski bed disturbance compared with small floods that mainly et al. 2017). Outside of these experimental studies in well generate turbidity events (Cronin et al. 2007; O’Connor protected watersheds, we have had limited success in uncov- et al. 2012). It is likely, yet untested, that the mechanism ering consistent relationships between nutrient supply and through which floods impact river productivity may shift in algal biomass across rivers (Francoeur et al. 1999; Dodds and larger rivers where a greater proportion of GPP may be per- Smith 2016). Since nutrient loading from farm or road run- formed by algae within the water column (Oliver and Mer- off and wastewaters is often coincident with changes in rick 2006). In these larger rivers, we may expect the light organic matter and sediment loading and changes in distur- attenuation caused by flood-associated pulses of suspended bance regimes, it is perhaps not surprising that differences in sediment that raise turbidity to become the dominant distur- these ultimate constraints (light and disturbance) override bance impact rather than bed disturbance. our ability to detect the additional effects of nutrient supply. Low- and no-flow drought conditions are also strong con- straints on river ecosystem productivity. The most conspicu- Linking river “climate” with river metabolism ous effect is obviously the reduction of total habitat area Though we know that light, temperature and hydrologic associated with drying (Stanley et al. 1997). Within the disturbance (a river’s “climate”) are primary determinants of shrinking wetted channel area, responses to diminishing metabolic activity in rivers, and that nutrients may acceler- flow can be highly variable. In one case, progression of dry- ate the metabolic response to each of these drivers, we have ing coupled with high light conditions were associated with only limited information with which to predict how the dis- growth of floating algae and extremely high rates of produc- tinct temporal dynamics of each driver interact to determine tivity in increasingly isolated pools (Acuna~ et al. 2005). river metabolic regimes. Because of extensive historic invest- Under such conditions, dissolved organic carbon can become ments in flow monitoring, we have large amounts of data highly concentrated. As streambed sediments are desiccated, available to understand fine scale variation in river flows primary production ceases in surface sediments and reestab- (Poff et al. 1997, 2006) and thermal regimes (Olden and lishment may require an extended period of recovery after Naiman 2009; Maheu et al. 2016), but we have far less infor- rewetting in some ecosystems (e.g., desert streams, Stanley mation about river light regimes (Hill 1996). We lack empiri- cal measurements of light availability to the river surface for all but the largest rivers and a few well studied streams (Hill 1First recorded in 1508 by Erasmus in his collection of Latin prov- and Dimick 2002; Hill et al. 2011). We have even less infor- erbs, Adagia. mation about the light regimes experienced by the primary

S105 Bernhardt et al. Metabolic regimes

Table 1. Changes in global temperatures and land cover will interact with widespread flow regulation to alter the light, thermal and flow regimes that ultimately shape the metabolic regimes of rivers. Direct effects of these global trends on the drivers of metabo- lism appear in black boxes while indirect effects appear as regular text.

producers attached to bed sediments or transported in river high light, high nutrients, and stable flow—the “windows of flow, and thus we have limited empirical data and few mod- opportunity” for high metabolic rates within stream ecosys- els with which to estimate light availability to river auto- tems—seem likely to shift in their timing, duration and mag- trophs (Ochs et al. 2013). Better information on this aspect nitude because of these competing drivers of change (e.g., of the “climate” of rivers is sorely needed to generate mecha- Ulseth et al. 2017). nistic models that can predict the metabolic capacity of river Allochthonous carbon and river metabolism ecosystems. Further complicating efforts to measure and model river Such model development is essential if we want to better metabolism is that the energetic basis of many river food understand how river ecosystems are being altered by three webs is not limited to in situ productivity. Streams with major trajectories of change: rising global temperatures; land dense canopies of riparian vegetation, rivers carrying high use change; and flow regulation (Table 1). Climate change is particulate loads, and blackwater rivers all can support leading to rising temperatures and altered hydrographs for diverse and rich food webs and high rates of ecosystem respi- rivers globally (Oki and Kanae 2006; Grimm et al. 2013). ration that are dominantly or exclusively supported by car- Land use change is increasing sediment loads, nutrient bon derived from upslope and upstream ecosystems (Fisher inputs and flood and drought frequency for at least some and Likens 1973; Meyer et al. 1997; Wallace et al. 1997; portion of the channel network in every major river basin Moore et al. 2004). At the other end of the spectrum, rivers (Vorosmarty and Sahagian 2000; Vorosmarty et al. 2000; with high water clarity and no canopy shading still receive Allan 2004). The widespread construction of dams (Lehner subsidies to the autochthonous-based food web in the form et al. 2011; Zarfl et al. 2014) has fundamentally altered the of particulate detritus and dissolved organic matter from the flow regimes of many rivers (sensu Poff et al. 1997). The surrounding watershed. These fixed carbon subsidies are implications for river ecosystem energetics are difficult to often delivered to rivers in pulses. These may be regular and predict, in part because the direct impacts of these drivers predictable (e.g., annual litterfall or snowmelt fluxes of dis- on river metabolic rates can be antagonistic (higher solved organic matter); frequent but unpredictable (e.g., nutrients vs. higher disturbance frequency) and because inputs of large amounts of DOM during floods; Boyer et al. these direct impacts may be mitigated or enhanced by the 1997; Raymond and Saiers 2010); or very infrequent cata- response of riparian vegetation to climate change. Periods of strophic events (e.g., hurricanes, ice storms, fires) that may

S106 Bernhardt et al. Metabolic regimes

Fig. 4. A conceptual model of four possible and contrasting patterns of annual productivity that could arise for river ecosystems within the same lati- tude and terrestrial biome. The annual pattern of solar radiation (above any vegetation) is shown in yellow while GPP is shown in green. Maximum potential GPP is always constrained by incident light. Disturbance and the interception of light by canopy, DOC or sediment each reduce the capacity for river autotrophs to levels of productivity well below that predicted from light availability alone. deliver large quantities of terrestrial organic matter to river A global effort to measure, model and synthesize metabo- ecosystems (Bernhardt et al. 2003; Earl and Blinn 2003; lism across many river ecosystems will reveal controls on the Dahm et al. 2015). Floods deliver, bury and remove the rates, timing and sensitivity of freshwater metabolism. Dif- organic matter stored within riverbeds. The extent to which ferences in metabolic regimes among rivers, and changes in rivers store or transmit this supply of fixed OM downstream these regimes over time, are likely to have wide-ranging eco- will determine the degree to which ER is tied to temporal logical consequences. Metabolic regimes incorporate not variation in river productivity. In those river reaches where only the amount of energy available to fuel secondary pro- large standing stocks of organic matter accumulate, rates of duction, but also the temporal pattern of energy availability ER may be decoupled from GPP, and the stored OM can sup- that determines which consumers are phenologically best port significant biological activity in rivers where primary suited to capitalize on metabolic products. Understanding productivity is negligible (reviewed in Webster and Meyer ecosystem phenology and what it reveals about coupling 1997). Differences between rivers in their capacity to store between energetics and element cycling will open new OM introduces considerable variation in the extent to which opportunities to apply our emerging understanding of meta- autochthonous and allochthonous energy sources support bolic regimes as diagnostic tools for river management. river food webs, as well as the coherence between seasonal Three frontier opportunities suggest themselves as particu- and disturbance recovery patterns of GPP and ER. larly exciting avenues for progress: Annual patterns of river metabolism As we amass a new understanding of the annual patterns FRONTIER # 1 – linking metabolic regimes and of metabolism in rivers, we anticipate that there are likely to organismal phenology be discrete classes of river ecosystems that share a character- Ecosystem rates of GPP and ER are maximized at times istic rhythm in the foundational metabolic processes of GPP when and places where multiple potentially limiting resour- and ER (e.g., Fig. 4). These “metabolic regimes” may be char- ces are abundant and the biomass of primary producers is at acterized by differences in: mean rates and their variability its peak. Variation in resource supply and biomass removal and skewness; the temporal structure of metabolism such as thus drive considerable temporal variation in energy produc- the autocorrelation and periodicity of ER and GP; the dura- tion across systems (Fig. 5). The study of the match or mis- tion and timing of metabolic peaks and troughs; the post- match between an organism’s life history and organic matter disturbance rates of recovery; and by the variation in these supply has a rich history in stream ecology, where many characteristics among years and decades. These patterns are consumers are dependent upon relatively discrete and pre- likely to be linked to the overlapping temporal patterns of dictable periods of either autochthonous production (e.g., the many factors that potentially control metabolism, and spring algal blooms) or allochthonous organic matter supply thus provide diagnostic information to determine the pri- (e.g., autumn litter fall). Disturbances that occur within mary constraints on metabolism and to predict the likely these short periods of peak metabolic activity may have dis- energetic consequences of climate change, land use change proportionate effects on annual rates of ecosystem metabo- or flow management for river ecosystems. Particularly for lism as well as on the productivity and identity of secondary river ecosystems that have only short periods of peak metab- consumers (Wootton et al. 1996; Power et al. 2008; Spon- olism, small shifts in the timing or magnitude of resource seller et al. 2010). In some rivers, changes in the identity of supply or of the peak flows, droughts or other events that secondary consumers associated with intermediate levels of constrain or reset metabolic rates may have disproportion- disturbance can lead to increased production of tertiary and ately large impacts on river productivity at daily, seasonal, higher trophic levels even as primary and secondary con- and annual time scales. sumption decline (Power et al. 1996; Cross et al. 2011).

S107 Bernhardt et al. Metabolic regimes

However, more frequent disturbances have been shown and are predicted to favor small bodied, fast growing consumers (Gray 1981; Fisher and Gray 1983). Increased mortality and longer recovery times of larger predatory animals are pro- posed explanations for this trend. It is also possible, and likely, that reductions in the total energy available to con- sumers because of the removal of autotrophs and organic matter are in part responsible for these negative effects on higher trophic levels (as per predictions of the trophic pyra- mid, Elton 1927). We suggest that simultaneous measurement and model- ing of river ecosystem energetics and food web properties will be needed to predict the likely consequences of climate and land use change (Table 1) on river biodiversity and food web structure. Sabo et al. (2010) documented a negative rela- tionship between flow variability (estimated from 16 to 20 yr of continuously measured flow data) and food chain length (FCL) for 20 U.S. rivers. In the same study, the authors con- cluded that there was no relationship between FCL and rates of gross primary production (Sabo et al. 2010). Since the GPP estimates used in the Sabo et al. (2010) synthesis were derived from only 1 to 20 d of midsummer (low flow) meas- urements they are likely highly inaccurate estimates of annual ecosystem productivity. As our measurements of river metabolism grow more similar in temporal resolution and longevity to our hydrologic records we predict we will dis- cover strong, synergistic interactions between flow and meta- bolic regimes in constraining biodiversity and food web structure. There is mounting evidence that frequent high flows significantly constrain river GPP (Uehlinger 2006; Beaulieu et al. 2013, Fig. 3), and we suggest that the altera- tions in animal assemblages often attributed to hydrologic alteration or disturbance (Poff and Ward 1989; Sabo et al. 2009; Poff and Zimmerman 2010) may be due in part to the indirect effects of these flow alterations on ecosystem metab- olism. If more comprehensive data support this prediction, it will only reinforce the importance of maintaining and restoring natural flow regimes to sustain both the distur- bance and the energetic regimes that structure river food webs. Widespread measurement of river ecosystem metabolism offers a new opportunity to confront this rich conceptual understanding with large volumes of empirical data describ- ing the seasonal patterns of energy production and dissipa- tion in rivers (Fig. 5). Coupling ecosystem metabolism time series with contemporaneous measurements of benthic sec- ondary production or food web characteristics is likely to lead to new insights and re-examination of current assump- Fig. 5. Cumulative annual metabolism graphs for the four rivers from tions. For example, studies of the impacts of altered flow 22 Fig. 1. GPP is in green, ER is in brown. Units on the y axis are in g O2 m . regimes on the species composition of insect and fish com- Note that the southern river, Five Mile Creek has relatively uniform GPP while the northern Menominee River has high accumulation only during munities often focus on how the life history of declining or the summer months. The San Antonio River in Texas has alternating peri- extirpated taxa depended on natural flow regimes (Lytle and ods of high and very low productivity, while Fanno Creek has low GPP Poff 2004; Lytle et al. 2008). There has been far less consider- year-round. Rates of ER exceed GPP annually and for most daily time steps ation of how the direct impacts of flow modification on in all four streams. S108 Bernhardt et al. Metabolic regimes higher trophic levels may be comfounded by the indirect expanding portfolio of constituents continuously over long effects of flow on metabolism, the energetic base of these time periods (Porter et al. 2012; Pellerin et al. 2016). food webs. A cursory look at newly available data from just a Diel oscillations in a wide range of solutes are common in few rivers in the United States (e.g., Fig. 3) demonstrates rivers and are linked to metabolism by both direct and indi- that high flows consistently reduce the biomass and cumula- rect mechanisms (Nimick et al. 2011; Hensley and Cohen tive GPP of river autotrophs. It is not just the “green,” or 2016). By quantifying variation in diel oscillations over time, algal supported food web that is likely to be affected. Many recent empirical studies have demonstrated close coupling river biota consume detritus delivered from the surrounding between instantaneous rates of metabolism and the assimila- terrestrial ecosystem, with the growth and diversity of shred- tion of dissolved nutrients (e.g., Heffernan and Cohen 2010; ders and detritivores strongly tied to the standing stocks of Rode et al. 2016). While links between metabolism and leaf litter in small streams (Wallace et al. 1997, 1999). In soil nutrient cycling are also observed across rivers (e.g., Hall and and sediment food webs, heterotrophs rely on slow turnover Tank 2003), high-frequency continuous monitoring can illu- of massive standing stocks of organic carbon—but in streams minate relationships between metabolic processes and and rivers this “brown food web” is far more temporally longer-term changes in material export. For example, in a dynamic due to events that can export, bury or fragment Tennessee stream, daily GPP can drive variation in both diel NO2 oscillations and total uptake (Roberts and Mulholland benthic organic matter standing stocks (Wallace et al. 1995). 3 2007) (Fig. 6); similarly, interannual variation in springtime Thus, shifts in the timing of OM supply (via changes in ter- GPP due to storm frequency and timing accounts for much restrial phenology) and removal or burial (via changes in of interannual variation in watershed nutrient export (Lutz storm magnitudes or timing) may have important implica- et al. 2012) (Fig. 6). tions for the wide variety of shredders and collectors that Joint high-frequency measurements of oxygen and other rely on these allochthonous carbon sources. The lost produc- solutes can also illuminate the cascade of interactions tion that results from increasingly frequent or severe flows between metabolism and other biogeochemical processes. In will certainly outlast the impact of the physical disturbance a spring-fed Florida River, variation in metabolism drives diel and, in some systems, may substantially alter the commu- and seasonal variation in NO2 uptake, and also directly nity of secondary consumers, with additional effects at 3 affects day-to-day variation in denitrification, which higher trophic levels. Shifts in peak and total annual produc- 2 accounts for the majority of riverine NO3 removal (Hef- tivity resulting from global change drivers (Table 1) are likely fernan and Cohen 2010). In the same system, metabolism to have important implications for the biomass and diversity 32 affects diel variation in PO4 concentrations, both directly of aquatic consumers that can be supported within a river, through assimilatory uptake and indirectly via effects of pH while shifts in the timing of peak productivity (Table 1) may on calcite saturation (de Montety et al. 2011; Cohen et al. favor new types of animals at the expense of historically 2013). dominant taxa. Riverine metabolism is obviously not the only potential cause of fine-scale variation in river chemistry; indeed, many FRONTIER # 2 - coupling energetics and element such patterns reflect hydrologic and anthropogenic processes cycling occurring in hillslopes, riparian zones, and upstream within river networks (Kirchner et al. 2001; Harvey 2016). For exam- In rivers, indeed in all ecosystems, cycling of nutrients ple, daily variation in evapotranspiration rates of catchment and other elements can both depend on and influence meta- vegetation can drive diel variation in streamflow and thus bolic patterns, and these reciprocal interactions operate over the relative contribution of groundwater and shallow subsur- a wide range of time scales, within days, across seasons and face flowpaths, a mechanism that can drive diel variation in through recovery from antecedent disturbances (Kirchner stream solute concentrations even in the absence of high et al. 2004; Appling and Heffernan 2014). There is a rich his- rates of instream nutrient assimilation. Assimilation and tory of stream ecological research that has documented transformation processes well upstream of a sampling loca- strong links between metabolic phenology and the timing tion can result in temporal variation in solute concentra- and magnitude of nutrient retention (Meyer 1980; Mulhol- tions which are inconsistent in their timing or magnitude land et al. 1985; Meyer et al. 1998; Roberts and Mulholland with local control by in-stream processes (Pellerin et al. 2007). The emerging transition of stream metabolism studies 2009, 2014). Relatively few studies have attempted to disen- from daily rates to annual regimes is part of a much wider tangle hydrologic and biotic drivers of diel nutrient variation effort to document and draw inference from hydrochemical using direct measurements of both processes (but see Aubert patterns across the full span of relevant time scales, from and Breuer 2016; Lupon et al. 2016). days and storm events to seasons and decades (Kirchner Although nutrients do not appear to drive major differ- et al. 2004). These efforts are greatly enabled by new tech- ences in metabolism across rivers, it is likely that nutrients nologies and approaches that can measure a rapidly do limit metabolism in rivers where nutrient supply is low,

S109 Bernhardt et al. Metabolic regimes

Fig. 6. Data collected from Walker Branch watershed in Oak Ridge Tennessee (Roberts and Mulholland 2007; Roberts et al. 2007; Lutz et al. 2012). Upper panels during periods of the year with high GPP, diel increases in dissolved oxygen are linked to diel declines in stream nitrate concentrations while nitrate concentrations are constant (and elevated) during summer months when the forest canopy is closed. Middle panels: Seasonal variation in GPP explains the low stream nitrate concentrations during spring and fall. Lower panels: Years with larger spring algal blooms (e.g., 2005) have greater total annual instream nitrogen uptake than years in which spring storms reduce the longevity and peak productivity of spring algae (e.g., 2006). light is abundant and disturbance is infrequent. It has Taking full advantage of the growing capabilities of in been suggested that continuous daily estimates of nutrient situ sensors will require ongoing methodological and theo- concentrations could provide strong direct evidence for retical advances (Pellerin et al. 2016; Rode et al. 2016). nutrient dependence of metabolism (Hall 2016). At still Within rivers, the physical dynamics of transport are likely finer scales of diel variation, theoretical models predict to affect covariation of solutes in ways that may both mask that the magnitude of diel variation, relative to autotro- and influence biogeochemical processes. As one example, phic demand, may depend on whether nutrients are avail- the footprint of processes that involve dissolved gases is able in sufficient supply (Appling and Heffernan 2014). likely to differ from those that lack gas phase reactants and High-resolution sensors are particularly valuable for testing products (Hensley and Cohen 2016). Similar challenges may such links between organismal physiology and ecosystem arise in linking and disentangling processes that act at very processes. different rates or in opposition (e.g., nitrification vs.

S110 Bernhardt et al. Metabolic regimes

Fig. 7. The metabolic fingerprint is a diagnostic tool for comparing annual patterns of metabolism across rivers or across years for the same river. Panel A shows the metabolic fingerprint for 2014 GPP and ER estimates for each of the four streams shown in Fig. 1. The “fingerprint” is represented as a kernel density plot of all daily estimates of GPP and ER rates observed within each stream. The area of a river’s fingerprint (its variation along both the 1 : 1 line and ER axis) should be compressed by hydrologic disturbances and expanded by the supply of either solar energy of fixed carbon.

denitrification) within different habitat compartments (Hel- To take advantage of this new information, we need new ton et al. 2012; Gomez-Velez and Harvey 2014). Experimen- diagnostic tools that are easy to explain and interpret. We tal approaches that take advantage of sensor capabilities may propose the metabolic fingerprint as one such diagnostic help provide estimates of whole system and compartment- tool and framework for hypothesis testing (Fig. 7). The meta- specific biogeochemical processes. bolic fingerprint can be represented as the entire distribution One clear need is the further development of integrative, of daily estimates of GPP and ER that are observed for a ecosystem-level models that link metabolic, biogeochemical, river, or the summary of those data into kernel density plots and hydrologic processes within rivers. Widely used chemi- that allow easy visualization of both peak and median meta- cal loading models (i.e., Sparrow, HYDRA) couple dynamic bolic rates as well as variance in their ratio (Fig. 7). In Fig. 7, hydrology with static assumptions about element processing we show an explanatory cartoon of the fingerprint graph within rivers. At present, river science lacks the portfolio of alongside the “fingerprints” of the four rivers presented pre- theoretical and modeling approaches akin to those devel- viously in Figs. 1, 3. The highly productive Menominee River oped by terrestrial biogeochemists and ecologists (e.g., CEN- in Wisconsin occupies a much larger volume of metabolic TURY, Biome-BGC, P-NET) that allow for dynamic, non- space than does the very frequently flooded Fanno Creek in equilibrium modeling of ecosystem and biogeochemical pro- Oregon. We hypothesize that higher light and nutrient sup- cesses. Because rivers provide opportunities to observe meta- plies will expand both the total area and the maximal rates bolic and biogeochemical processes over comparable time represented by a river’s fingerprint, while hydrologic distur- scales, novel river ecosystem models offer the potential to bance and sediment loading will constrain the metabolic fin- develop and test more general theory of the interactions gerprint to values near the origins of both the GPP and ER among element cycles and energy flow in ecological systems. axes. We suspect that as we gain annual metabolism data from an increasing number of rivers we will discover recog- Frontier # 3 - river metabolism for diagnosis and nizable clusters of river metabolic regimes. As we develop management our mechanistic understanding of how river metabolic Basal metabolism constrains the “work” that ecosystems regimes are established, maintained and disrupted, we will can do and we can measure this process for an entire ecosys- become better able to determine how river metabolic tem just as we do for an organism through monitoring its regimes are being altered by pollution, flow regulation or cli- exchange of O2 or CO2 with the atmosphere. Many govern- mate change as well as how effective mitigation, restoration, ment agencies monitor dissolved oxygen in freshwater eco- and preservation efforts are at returning rivers toward more systems to avoid or regulate low oxygen events. At no natural metabolic regimes. additional costs, these same data can be repurposed to pro- In addition to comparing fingerprints across streams, vide real time, continuous measures of ecosystem function. these same visualization tools can be used to compare across

S111 Bernhardt et al. Metabolic regimes

Fig. 8. Two photos of the Oria River and its metabolic fingerprint before (black) and after (green) the construction of a wastewater treatment plant. Photo credits to M. Arroita and A. Elosegi. years or among sites within the same river to examine the distribution of metabolic rates—the “metabolic finger- effects of infrastructure or restoration efforts, land use print”—of the Oria River (Fig. 8). Daily amplitudes of dis- change, or hydrologic extremes. For example, the metabolic solved oxygen greatly decreased due to the reduced fingerprint of the Oria River in the northern Spain changed metabolic rates with oxygen saturation now consistently in response to the implementation of a wastewater treatment between 80% and 120%, resulting in greatly improved con- plant (WWTP) (Fig. 8). The Oria River drains a basin of ditions for river biota. This case study shows just one exam- 2 882 km with 126,000 inhabitants in which industries had ple of how the metabolic “fingerprint” can serve as both a previously diverted their wastewater directly to the river. sensitive detector and a clear outreach tool to describe eco- Historic water quality data collected by the monitoring sta- system functional responses to intervention. tion located in the upper basin ( 333 km2) revealed that untreated sewage directly coming from houses and industries Concluding remarks carried large amounts of both inorganic and organic pollu- tants, elevating ammonium in the river to concentrations as Syntheses of daily metabolism measurements from across highs as 2.5 mgL21. High loads of organic matter resulted in different rivers have revealed enormous variability and had very high rates of stream ER that were not accompanied by limited success in explaining this variance. We suggest that similarly high rates of GPP. Although hyperoxic conditions estimating GPP and ER for one or a few days provides little (saturation 150%) were common during the day due to the information about the energetics of most rivers where fre- high rates of GPP, even higher ER rates derived from organic quent disturbance and highly variable light regimes lead to pollution frequently resulted in hypoxia at night (satu- high temporal variation in GPP and ER rates. By examining ration < 40%). In 2003, a WWTP was put into operation the patterns of metabolism over entire years, we can observe approximately 8 km upstream from the water quality moni- how the extrinsic controls of light, heat, allochthonous toring station, and now treats the domestic sewage of 60,000 inputs, and disturbance together shape metabolism, and we inhabitants as well as industrial sewage. The amount of inor- can begin to understand and predict how these drivers have ganic and organic pollutants entering the river subsequently changed and are likely to change because of widespread flow declined, with average ammonium concentrations falling to regulation, climate change, land use, and eutrophication. 0.18 mgL21. This improvement in water quality resulted in Land use and climate change are altering the light, ther- reduced rates of both GPP and ER, significantly changing the mal and disturbance regimes of rivers at both local and

S112 Bernhardt et al. Metabolic regimes global scales. Climate change has already shifted the timing, Allan, J. D. 2004. Landscapes and riverscapes: The influence magnitude and frequency of flooding events in rivers of land use on stream ecosystems. Annu. Rev. Ecol. Evol. throughout the world (Milly et al. 2002, 2005; Oki and Syst. 35: 257–284. doi:10.1146/annurev.ecolsys.35. Kanae 2006). Even in the absence of directional climate 120202.110122 change, hydropower and water supply dams are leading to Appling, A. P., and J. B. Heffernan. 2014. Nutrient limitation major shifts in the hydrologic regimes of most the world’s and physiology mediate the fine-scale (de)coupling of bio- rivers (Poff and Zimmerman 2010; Vor€ osmarty€ et al. 2010); geochemical cycles. Am. Nat. 184: 384–406. doi:10.1086/ while the frequency and magnitude of peak flows in smaller 677282 streams have been increased by urbanization’s spreading Atkinson, B. L., M. R. Grace, B. T. Hart, and K. E. N. influence (Paul and Meyer 2001; Walsh et al. 2005). All Vanderkruk. 2008. Sediment instability affects the rate forms of land use change tend to generate increased turbid- and location of primary production and respiration in a ity events and accelerated bed and bank erosion (Leopold sand-bed stream. J. North Am. Benthol. Soc. 27: 581–592. et al. 1964). Approaches to manage the stressors and protect doi:10.1899/07-143.1 ecosystem services of rivers are in their infancy, and we sug- Aubert, A. H., and L. Breuer. 2016. New seasonal shift in in- gest that effective river conservation and management stream diurnal nitrate cycles identified by mining high- requires that we devote as much attention to the metabolic frequency data. PloS One 11: e0153138. doi:10.1371/ regimes of rivers as we have already invested in understand- journal.pone.0153138 ing river flow regimes. Battin, T. J., S. Luyssaert, L. A. Kaplan, A. K. Aufdenkampe, Fortunately, we now have the technical and logistical A. Richter, and L. J. Tranvik. 2009. The boundless carbon capacity to measure dissolved gas and solute fluxes and cycle. Nat. Geosci. 2: 598–600. doi:10.1038/ngeo618 model ecosystem metabolism at the same highly resolved Beaulieu, J. J., C. P. Arango, D. A. Balz, and W. D. Shuster. time steps at which streamflow has historically been mea- 2013. Continuous monitoring reveals multiple controls on ecosystem metabolism in a suburban stream. Freshw. sured. If we can match these technological innovations with Biol. 58: 918–937. doi:10.1111/fwb.12097 equally innovative theoretical breakthroughs we will enable Berger, S. A., S. Diehl, H. Stibor, G. Trommer, and M. river ecosystem science to link energetics and element cycles Ruhenstroth. 2010. Water temperature and stratification at the time scales at which: organisms assimilate elements depth independently shift cardinal events during plank- (seconds to hours to days); disturbance initiates succession ton spring succession. Glob. Chang. Biol. 16: 1954–1965. (weeks to months); and terrestrial and aquatic element cycles doi:10.1111/j.1365-2486.2009.02134.x are linked (from hours to seasons). The emergent research Bernhardt, E., G. Likens, D. Buso, and C. Driscoll. 2003. In- will generate novel and sophisticated understanding of the stream uptake dampens effects of major forest disturbance multi-scale controls of stream metabolism, facilitating the on watershed nitrogen export. Proc. Natl. Acad. Sci. USA transition from short-term, reach-scale studies toward annu- 100: 10304–10308. doi:10.1073/pnas.1233676100 alized, scalable models of freshwater ecosystem energetics. In Bernot, M. J., and others. 2010. Inter-regional comparison of turn, these models will enable sophisticated predictions land-use effects on stream metabolism. Freshw. Biol. 55: about how river food webs, biodiversity, and nutrient proc- 1874–1890. doi:10.1111/j.1365-2427.2010.02422.x essing capacity are likely to change under realistic scenarios Biggs, B. J. F., R. A. Smith, and M. J. Duncan. 1999. Velocity of climate and land use change. and sediment disturbance of periphyton in headwater streams: Biomass and metabolism. J. North Am. Benthol. References Soc. 18: 222–241. doi:10.2307/1468462 Acuna,~ V., A. Giorgi, I. Munoz, U. Uehlinger, and S. Sabater. Bott, T. L., J. T. Brock, C. S. Dunn, R. J. Naimann, R. W. 2004. Flow extremes and benthic organic matter shape Ovink, and R. C. Peterson. 1985. Benthic community the metabolism of a headwater Mediterranean stream. metabolism in four temperate stream systems: An inter- Freshw. Biol. 49: 960–971. doi:10.1111/j.1365- biome comparison and evaluation of the river continuum 2427.2004.01239.x concept. Hydrobiologia 123: 3–45. doi:10.1007/ Acuna,~ V., I. Munoz, A. Giorgi, M. Omella, F. Sabater, and S. BF00006613 Sabater. 2005. Drought and postdrought recovery cycles Boyer, E. W., G. M. Hornberger, K. E. Bencala, and D. M. in an intermittent Mediterranean stream: Structural and McKnight. 1997. Response characteristics of DOC flushing functional aspects. J. North Am. Benthol. Soc. 24: 919– in an alpine catchment. Hydrol. Process. 11: 1635–1647. 933. doi:10.1899/04-078.1 doi:10.1002/(SICI)1099-1085(19971015)11:12 < 1635::AID- Acuna,~ V., C. Vilches, and A. Giorgi. 2011. As productive HYP494 >3.0.CO;2-H and slow as a stream can be-the metabolism of a Pampean Cohen, M. J., M. J. Kurz, J. B. Heffernan, J. B. Martin, R. L. stream. J. North Am. Benthol. Soc. 30: 71–83. doi: Douglass, C. R. Foster, and R. G. Thomas. 2013. Diel 10.1899/09-082.1 phosphorus variation and the stoichiometry of ecosystem

S113 Bernhardt et al. Metabolic regimes

metabolism in a large spring-fed river. Ecol. Monogr. 83: Finlay, J. C. 2011. Stream size and human influences on eco- 155–176. doi:10.1890/12-1497.1 system production in river networks. Ecosphere 2: art87. Cole, J. J., and others. 2007. Plumbing the global carbon cycle: doi:10.1890/ES11-00071.1 Integrating inland waters into the terrestrial carbon budget. Fisher, S. G., and G. E. Likens. 1973. Energy flow in Bear Ecosystems 10: 171–184. doi:10.1007/s10021-006-9013-8 Brook, New Hampshire - integrative approach to stream Cronin, G., J. H. J. McCutchan, J. Pitlick, and W. M. J. ecosystem metabolism. Ecol. Monogr. 43: 421–439. doi: Lewis. 2007. Use of Shields stress to reconstruct and fore- 10.2307/1942301 cast changes in river metabolism. Freshw. Biol. 52: 1587– Fisher, S. G., and L. J. Gray. 1983. Secondary production and 1601. doi:10.1111/j.1365-2427.2007.01790.x organic-matter processing by collector macroinvertebrates Cross, W. F., C. V. Baxter, K. C. Donner, E. J. Rosi-Marshall, in a desert stream. Ecology 64: 1217–1224. doi:10.2307/ T. A. Kennedy, R. O. Hall, H. A. W. Kelly, and R. S. 1937830 Rogers. 2011. Ecosystem ecology meets adaptive manage- Francoeur, S. N., B. J. F. Biggs, R. A. Smith, and R. L. Lowe. ment: Food web response to a controlled flood on the 1999. Nutrient limitation of algal biomass accrual in Colorado River, Glen Canyon. Ecol. Appl. 21: 2016–2033. streams: Seasonal patterns and a comparison of methods. doi:10.1890/10-1719.1 J. North Am. Benthol. Soc. 18: 242–260. doi:10.2307/ Dahm, C. N., R. I. Candelaria-Ley, C. S. Reale, J. K. Reale, 1468463 and D. J. V. Horn. 2015. Extreme water quality degrada- Gomez-Velez, J. D., and J. W. Harvey. 2014. A hydrogeomor- tion following a catastrophic forest fire. Freshw. Biol. 60: phic river network model predicts where and why hypo- 2584–2599. doi:10.1111/fwb.12548 rheic exchange is important in large basins. Geophys. Res. Davies-Colley, R. J., and D. G. Smith. 2001. Turbidity, sus- Lett. 41: 6403–6412. doi:10.1002/2014GL061099 pended sediment, and water clarity: A review. J. Am. Gray, L. J. 1981. Species composition and life histories of Water Resour. Assoc. 37: 1085–1101. doi:10.1111/j.1752- aquatic insects in a lowland sonoran desert stream. Am. 1688.2001.tb03624.x Midl. Nat. 106: 229–242. doi:10.2307/2425159 de Montety, V., J. B. Martin, M. J. Cohen, C. Foster, and M. J. Grimm, N. B. 1987. Nitrogen dynamics during succession in Kurz. 2011. Influence of diel biogeochemical cycles on car- a desert stream. Ecology 68: 1157–1170. doi:10.2307/ bonate equilibrium in a karst river. Chem. Geol. 283:31–43. 1939200 doi:10.1016/j.chemgeo.2010.12.025 Grimm, N. B., and S. G. Fisher. 1986. Nitrogen limitation in Demars, B. O. L., J. R. Manson, J. S. Olafsson, G. M. a Sonoran Desert Stream. J. North Am. Benthol. Soc. 5:2– Gıslason, and N. Friberg. 2011a. Stream hydraulics and 15. doi:10.2307/1467743 temperature determine the metabolism of geothermal Ice- Grimm, N. B., and S. G. Fisher. 1989. Stability of periphyton landic streams. Knowl. Manag. Aquat. Ecosyst. 402: 05. and macroinvertebrates to disturbance by flash floods in a Demars, B. O. L., and others. 2011b. Temperature and the desert stream. J. North Am. Benthol. Soc. 8: 293–307. doi: metabolic balance of streams. Freshw. Biol. 56: 1106– 10.2307/1467493 1121. doi:10.1111/j.1365-2427.2010.02554.x Grimm, N. B., and others. 2003. Merging aquatic and terres- Dodds, W. K., and V. H. Smith. 2016. Nitrogen, phosphorus, trial perspectives of nutrient biogeochemistry. Oecologia and eutrophication in streams. Inland Waters 6: 155–164. 137: 485–501. doi:10.1007/s00442-003-1382-5 doi:10.5268/IW-6.2.909 Grimm, N. B., and others. 2013. The impacts of climate Earl, S. R., and D. W. Blinn. 2003. Effects of wildfire ash on change on ecosystem structure and function. Front. Ecol. water chemistry and biota in South-Western U.S.A. Environ. 11: 474–482. doi:10.1890/120282 streams. Freshw. Biol. 48: 1015–1030. doi:10.1046/j.1365- Hall, R. O., Jr. 2016. Metabolism of streams and rivers: Esti- 2427.2003.01066.x mation, controls, and applications, p. 151–180. In J. B. Elton, C. S. 1927. Animal ecology. Macmillan. Jones and E. H. Stanley [eds.], Stream ecosystems in a Fellows, C. S., J. E. Clapcott, J. W. Udy, S. E. Bunn, B. D. Harch, changing environment. Academic Press. M. J. Smith, and P. M. Davies. 2006. Benthic metabolism as Hall, R. O., and J. L. Tank. 2003. Ecosystem metabolism con- an indicator of stream ecosystem health. Hydrobiologia trols nitrogen uptake in streams in Grand Teton National 572: 71–87. doi:10.1007/s10750-005-9001-6 Park, Wyoming. Limnol. Oceanogr. 48: 1120–1128. doi: Field, C. B., M. J. Behrenfeld, J. T. Randerson, and P. 10.4319/lo.2003.48.3.1120 Falkowski. 1998. Primary production of the biosphere: Hall, R. O., C. B. Yackulic, T. A. Kennedy, M. D. Yard, E. J. Integrating terrestrial and oceanic components. Science Rosi-Marshall, N. Voichick, and K. E. Behn. 2015. Turbid- 281: 237–240. doi:10.1126/science.281.5374.237 ity, light, temperature, and hydropeaking control primary Fierer, N., and J. Schimel. 2002. A proposed mechanism for productivity in the Colorado River, Grand Canyon. Lim- the pulse in carbon dioxide production commonly nol. Oceanogr. 60: 512–526. doi:10.1002/lno.10031 observed following the rapid rewetting of a dry soil. Soil Harvey, J. W. 2016. Hydrologic exchange flows and their Sci. Soc. Am. J. 67: 798–805. doi:10.2136/sssaj2003.0798 ecological consequences in river corridors, p. 1–83. In J. B.

S114 Bernhardt et al. Metabolic regimes

Jones and E. H. Stanley [eds.], Stream ecosystems in a Julian, J. P., M. W. Doyle, S. M. Powers, E. H. Stanley, and J. changing environment. Academic Press. A. Riggsbee. 2008. Optical water quality in rivers. Water Heffernan, J. B., and M. J. Cohen. 2010. Direct and indirect Resour. Res. 44: W10411. doi:10.1029/2007WR006457 coupling of primary production and diel nitrate dynamics Kirchner, J. W., X. H. Feng, and C. Neal. 2001. Catchment- in a subtropical spring-fed river. Limnol. Oceanogr. 55: scale advection and dispersion as a mechanism for fractal 677–688. doi:10.4319/lo.2010.55.2.0677 scaling in stream tracer concentrations. J. Hydrol. 254: Helton, A. M., G. C. Poole, R. A. Payn, C. Izurieta, and J. A. 81–100. doi:10.1016/S0022-1694(01)00487-5 Stanford. 2012. Scaling flow path processes to fluvial land- Kirchner, J. W., X. Feng, C. Neal, and A. J. Robson. 2004. scapes: An integrated field and model assessment of tem- The fine structure of water-quality dynamics: The (high- perature and dissolved oxygen dynamics in a river- frequency) wave of the future. Hydrol. Process. 18: 1359. floodplain-aquifer system. J. Geophys. Res. Biogeosci. doi:10.1002/hyp.5537 117: 13. doi:10.1029/2012JG002025 Kominoski, J. S., A. D. Rosemond, J. P. Benstead, V. Gulis, Hensley, R. T., and M. J. Cohen. 2016. On the emergence of and D. W. P. Manning. 2017. Experimental nitrogen and diel solute signals in flowing waters. Water Resour. Res. phosphorus additions increase rates of stream ecosystem 52: 759–772. doi:10.1002/2015WR017895 respiration and carbon loss. Limnol. Oceanogr. 51: 639. Hill, W. R. 1996. Effects of light, p. 121–148. In B. Stevenson doi:10.1002/lno.10610 and R. L. Lowe [eds.] Algal ecology. Academic Press. Lamberti, G. A., and A. D. Steinman. 1997. A comparison of Hill, W. R., and S. M. Dimick. 2002. Effects of riparian leaf primary production in stream ecosystems. J. North Am. dynamics on periphyton photosynthesis and light utilisa- Benthol. Soc. 16: 95–104. doi:10.2307/1468241 tion efficiency. Freshw. Biol. 47: 1245–1256. doi:10.1046/ Lampert, W., W. Fleckner, H. Rai, and B. E. Taylor. 1986. j.1365-2427.2002.00837.x Phytoplankton control by grazing zooplankton – a study Hill, W. R., S. E. Fanta, and B. J. Roberts. 2009. Quantifying on the spring clear-water phase. Limnol. Oceanogr. 31: phosphorus and light effects in stream algae. Limnol. 478–490. doi:10.4319/lo.1986.31.3.0478 Oceanogr. 54: 368–380. doi:10.4319/lo.2009.54.1.0368 Lehner, B., and others. 2011. High-resolution mapping of Hill, W. R., B. J. Roberts, S. N. Francoeur, and S. E. Fanta. the world’s reservoirs and dams for sustainable river-flow 2011. Resource synergy in stream periphyton communi- management. Front. Ecol. Environ. 9: 494–502. doi: ties. J. Ecol. 99: 454–463. 10.1890/100125 Hoellein, T. J., D. A. Bruesewitz, and D. C. Richardson. 2013. Leith, H. 1975. Modeling the primary productivity of the Revisiting Odum (1956): A synthesis of aquatic ecosystem world, p. 237–263. In H. Leith and R. H. Whittaker [eds.], metabolism. Limnol. Oceanogr. 58: 2089–2100. doi: Primary productivity of the biosphere. Springer-Verlag. 10.4319/lo.2013.58.6.2089 Leopold, L. B., M. G. Wolman, and J. R. Miller. 1964. Fluvial Holtgrieve, G. W., D. E. Schindler, T. A. Branch, and Z. T. processes in geomorphology. W.H. Freeman. A’Mar. 2010. Simultaneous quantification of aquatic eco- Lewis, W. M., W. A. Wurtsbaugh, and H. W. Paerl. 2011. Ratio- system metabolism and reaeration using a Bayesian statis- nale for control of anthropogenic nitrogen and phosphorus tical model of oxygen dynamics. Limnol. Oceanogr. 55: to reduce eutrophication of inland waters. Environ. Sci. 1047–1063. doi:10.4319/lo.2010.55.3.1047 Technol. 45: 10300–10305. doi:10.1021/es202401p Hope, A. J., W. H. McDowell, and W. M. Wollheim. 2014. Lupon, A., E. Martı, F. Sabater, and S. Bernal. 2016. Green Ecosystem metabolism and nutrient uptake in an urban, light: Gross primary production influences seasonal piped headwater stream. Biogeochemistry 121: 167–187. stream N export by controlling fine-scale N dynamics. doi:10.1007/s10533-013-9900-y Ecology 97: 133–144. doi:10.1890/14-2296.1 Houser, J. N., P. J. Mulholland, and K. O. Maloney. 2005. Lutz, B. D., P. J. Mulholland, and E. S. Bernhardt. 2012. Catchment disturbance and stream metabolism: Patterns in Long-term data reveal patterns and controls on stream ecosystem respiration and gross primary production along a water chemistry in a forested stream: Walker Branch, Ten- gradient of upland soil and vegetation disturbance. J. North nessee. Ecol. Monogr. 82: 367–387. doi:10.1890/11-1129.1 Am. Benthol. Soc. 24: 538–552. doi:10.1899/04-034.1 Lytle, D. A., and N. L. Poff. 2004. Adaptation to natural flow Huryn, A. D., J. P. Benstead, and S. M. Parker. 2014. Seasonal regimes. Trends Ecol. Evol. 19: 94–100. doi:10.1016/ changes in light availability modify the temperature j.tree.2003.10.002 dependence of ecosystem metabolism in an arctic stream. Lytle, D. A., M. T. Bogan, and D. S. Finn. 2008. Evolution of Ecology 95: 2826–2839. doi:10.1890/13-1963.1 aquatic insect behaviours across a gradient of disturbance Izagirre, O., U. Agirre, M. Bermejo, J. Pozo, and A. Elosegi. predictability. Proc. R. Soc. B Biol. Sci. 275: 453–462. doi: 2008. Environmental controls of whole-stream metabo- 10.1098/rspb.2007.1157 lism identified from continuous monitoring of Basque Maheu, A., N. L. Poff, and A. St-Hilaire. 2016. A classification streams. J. North Am. Benthol. Soc. 27: 252–268. doi: of stream water temperature regimes in the conterminous 10.1899/07-022.1 USA. River Res. Appl. 32: 896–906. doi:10.1002/rra.2906

S115 Bernhardt et al. Metabolic regimes

McDowell, W. H. 2015. NEON and STREON: Opportunities aqueous chemistry of streams: A review. Chem. Geol. and challenges for the aquatic sciences. Freshw. Sci. 34: 283: 3–17. doi:10.1016/j.chemgeo.2010.08.017 386–391. doi:10.1086/679489 Noormets, A. [ed.]. 2009. Phenology of ecosystem processes: McKnight, D. M., C. M. Tate, E. D. Andrews, D. K. Niyogi, K. Applications in global change research, 275 p. Springer- Cozzetto, K. Welch, W. B. Lyons, and D. G. Capone. Verlag. 2007. Reactivation of a cryptobiotic stream ecosystem in Ochs, C. A., O. Pongruktham, and P. V. Zimba. 2013. Dark- the McMurdo Dry Valleys, Antarctica: A long-term geo- ness at the break of noon: Phytoplankton production in morphological experiment. Geomorphology 89: 186–204. the Lower Mississippi River. Limnol. Oceanogr. 58: 555– doi:10.1016/j.geomorph.2006.07.025 568. doi:10.4319/lo.2013.58.2.0555 McTammany, M. E., J. R. Webster, E. F. Benfield, and M. A. O’Connor, B. L., J. W. Harvey, and L. E. McPhillips. 2012. Neatrour. 2003. Longitudinal patterns of metabolism in a Thresholds of flow-induced bed disturbances and their southern Appalachian river. J. North Am. Benthol. Soc. effects on stream metabolism in an agricultural river. Water 22: 359–370. doi:10.2307/1468267 Resour. Res. 48: W08504. doi:10.1029/2011WR011488 Merbt, S. N., L. Proia, J. I. Prosser, E. Marti, E. O. Casamayor, Odum, H. T. 1956. Primary production in flowing waters. Lim- and D. von Schiller. 2016. Stream drying drives microbial nol. Oceanogr. 1: 102–117. doi:10.4319/lo.1956.1.2.0102 ammonia oxidation and first-flush nitrate export. Ecology Odum, H. T. 1957. Primary production measurements in 97: 2192–2198. doi:10.1002/ecy.1486 eleven Florida springs and a marine turtle-grass community. Meyer, J. L. 1980. Dynamics of phosphorus and organic mat- Limnol. Oceanogr. 2: 85–97. doi:10.4319/lo.1957.2.2.0085 ter during leaf decomposition in a forest stream. Oikos Oki, T., and S. Kanae. 2006. Global hydrological cycles and 34: 44–53. doi:10.2307/3544548 world water resources. Science 313: 1068–1072. doi: Meyer, J. L., A. C. Benke, R. T. Edwards, and J. B. Wallace. 10.1126/science.1128845 1997. Organic matter dynamics in the Ogeechee River, a Olden, J. D., and R. J. Naiman. 2009. Incorporating thermal blackwater river in Georgia, USA. J. North Am. Benthol. regimes into environmental flows assessments: Modifying Soc. 16: 82–87. doi:10.2307/1468239 dam operations to restore freshwater ecosystem Meyer, J. L., J. B. Wallaca, and S. L. Eggert. 1998. Leaf litter integrity. Freshw. Biol. 55: 86–107. doi:10.1111/j.1365- as a source of dissolved organic carbon to streams. Ecosys- 2427.2009.02179.x tems 1: 240–249. doi:10.1007/s100219900019 Oliver, R. L., and C. J. Merrick. 2006. Partitioning of river Milly, P. C. D., R. T. Wetherald, K. A. Dunne, and T. L. metabolism identifies phytoplankton as a major contribu- Delworth. 2002. Increasing risk of great floods in a chang- tor in the regulated Murray River (Australia). Freshw. Biol. ing climate. Nature 415: 514–517. doi:10.1038/415514a 51: 1131–1148. doi:10.1111/j.1365-2427.2006.01562.x Milly, P. C. D., K. A. Dunne, and A. V. Vecchia. 2005. Global Paul, M. J., and J. L. Meyer. 2001. Streams in the urban land- pattern of trends in streamflow and water availability in a scape. Annu. Rev. Ecol. Syst. 32: 333–365. doi:10.1146/ changing climate. Nature 438: 347–350. doi:10.1038/ annurev.ecolsys.32.081501.114040 nature04312 Pellerin, B. A., B. D. Downing, C. Kendall, R. A. Dahlgren, T. Minshall, G. W., R. C. Petersen, T. L. Bott, C. E. Cushing, K. E. C. Kraus, J. Saraceno, R. G. M. Spencer, and B. A. W. Cummins, R. L. Vannote, and J. R. Sedell. 1992. Bergamaschi. 2009. Assessing the sources and magnitude Stream ecosystem dynamics of the Salmon River, Idaho - of diurnal nitrate variability in the San Joaquin River an 8th-order system. J. North Am. Benthol. Soc. 11: 111– (California) with an in situ optical nitrate sensor and dual 137. doi:10.2307/1467380 nitrate isotopes. Freshw. Biol. 54: 376–387. doi:10.1111/ Moore, J. C., and others. 2004. Detritus, trophic dynamics j.1365-2427.2008.02111.x and biodiversity. Ecol. Lett. 7: 584–600. doi:10.1111/ Pellerin, B. A., B. A. Bergamaschi, R. J. Gilliom, C. G. j.1461-0248.2004.00606.x Crawford, J. Saraceno, C. P. Frederick, B. D. Downing, and Mulholland, P. J., J. D. Newbold, J. W. Elwood, and J. R. J. C. Murphy. 2014. Mississippi River nitrate loads from Webster. 1985. Phosphorus spiralling in a woodland high frequency sensor measurements and regression-based stream: Seasonal variations. Ecology 66: 1012–1023. doi: load estimation. Environ. Sci. Technol. 48: 12612–12619. 10.2307/1940562 doi:10.1021/es504029c Mulholland, P. J., and others. 2001. Inter-biome comparison Pellerin, B. A., B. A. Stauffer, D. A. Young, D. J. Sullivan, S. of factors controlling stream metabolism. Freshw. Biol. B. Bricker, M. R. Walbridge, G. A. Clyde, and D. M. Shaw. 46: 1503–1517. doi:10.1046/j.1365-2427.2001.00773.x 2016. Emerging tools for continuous nutrient monitoring Mulholland, P. J., and others. 2008. Stream denitrification networks: Sensors advancing science and water resources across biomes and its response to anthropogenic nitrate protection. J. Am. Water Resour. Assoc. 52: 993–1008. loading. Nature 452: 202–205. doi:10.1038/nature06686 doi:10.1111/1752-1688.12386 Nimick, D.A.N.D.A., C. H. Gammons, and S. R. Parker. 2011. Peterson, B. J., and others. 1985. Transformation of a tundra Diel biogeochemical processes and their effect on the river from heterotrophy to autotrophy by addition of

S116 Bernhardt et al. Metabolic regimes

phosphorus. Science 229: 1383–1386. doi:10.1126/ Roley, S. S., J. L. Tank, N. A. Griffiths, and R. O. Hall, Jr. science.229.4720.1383 2014. The influence of floodplain restoration on whole- Poff, N. L., and J. V. Ward. 1989. Implications of streamflow stream metabolism in an agricultural stream: Insights variability and predictability for lotic community struc- from a 5-year continuous data set. Freshw. Sci. 33: 1043– ture - a regional-analysis of streamflow patterns. Can. J. 1059. doi:10.1086/677767 Fish. Aquat. Sci. 46: 1805–1818. doi:10.1139/f89-228 Sabo, J. L., J. C. Finlay, and D. M. Post. 2009. Food chains in Poff, N. L., J. D. Allan, M. B. Bain, J. R. Karr, K. L. freshwaters, p. 187–220. In R. S. Ostfeld and W. H. Schle- Prestegaard, B. D. Richter, R. E. Sparks, and J. C. singer [eds.], The Year in Ecology and Conservation Biol- Stromberg. 1997. The natural flow regime. Bioscience 47: ogy, Ann. N.Y. Acad. Sci. 1162: 187–220. doi:10.1111/ 769–784. doi:10.2307/1313099 j.1749-6632.2009.04445.x Poff, N. L., J. D. Olden, D. M. Pepin, and B. P. Bledsoe. 2006. Sabo, J. L., J. C. Finlay, T. Kennedy, and D. M. Post. 2010. Placing global stream flow variability in geographic and The role of discharge variation in scaling of drainage area geomorphic contexts. River Res. Appl. 22: 149–166. doi: and food chain length in rivers. Science 330: 965–967. 10.1002/rra.902 doi:10.1126/science.1196005 Poff, N. L., and J. K. H. Zimmerman. 2010. Ecological Schindler, D. W. 1978. Factors regulating phytoplankton pro- responses to altered flow regimes: A literature review to duction and standing crop in the world’s freshwaters. Lim- inform the science and management of environmental nol. Oceanogr. 23: 478–486. doi:10.4319/lo.1978.23.3.0478 flows. Freshw. Biol. 55: 194–205. doi:10.1111/j.1365- Solomon, C. T., and others. 2013. Ecosystem respiration: 2427.2009.02272.x Drivers of daily variability and background respiration in Poole, G. C., and C. H. Berman. 2001. An ecological perspec- lakes around the globe. Limnol. Oceanogr. 58: 849–866. tive on in-stream temperature: Natural heat dynamics and doi:10.4319/lo.2013.58.3.0849 mechanisms of human-caused thermal degradation. Envi- Sponseller, R. A., N. B. Grimm, A. J. Boulton, and J. L. Sabo. ron. Manag. 27: 787–802. doi:10.1007/s002670010188 2010. Responses of macroinvertebrate communities to Porter, J. H., P. C. Hanson, and C.-C. Lin. 2012. Staying long-term flow variability in a Sonoran Desert stream. afloat in the sensor data deluge. Trends Ecol. Evol. 27: Glob. Chang. Biol. 16: 2891–2900. doi:10.1111/j.1365- 122–129. doi:10.1016/j.tree.2011.11.009 2486.2010.02200.x Power, M. E., M. S. Parker, and J. T. Wootton. 1996. Distur- Stanley, E. H., S. G. Fisher, and N. B. Grimm. 1997. Ecosys- bance and food chain length in rivers, p. 286–297. In G. tem expansion and contraction in streams. Bioscience 47: A. Polis and K. O. Winemiller [eds.], Food webs: Integra- 427–435. doi:10.2307/1313058 tion of patterns and dynamics. Chapman and Hall. Stanley, E. H., N. J. Casson, S. T. Christel, J. T. Crawford, L. Power, M. E., M. S. Parker, and W. E. Dietrich. 2008. Sea- C. Loken, and S. K. Oliver. 2016. The ecology of methane sonal reassembly of a river food web: Floods, droughts, in streams and rivers: patterns, controls, and global signif- and impacts of fish. Ecol. Monogr. 78: 263–282. doi: icance. Ecol. Monogr. 86: 146–171. doi:10.1890/15-1027 10.1890/06-0902.1 Strayer, D. L., and D. Dudgeon. 2010. Freshwater biodiversity Raymond, P. A., and J. E. Saiers. 2010. Event controlled DOC conservation: Recent progress and future challenges. J. export from forested watersheds. Biogeochemistry 100: North Am. Benthol. Soc. 29: 344–358. doi:10.1899/08- 197–209. doi:10.1007/s10533-010-9416-7 171.1 Raymond, P. A., and others. 2013. Global carbon dioxide Tanentzap, A. J., and others. 2017. Terrestrial support of lake emissions from inland waters. Nature 503: 355–359. doi: food webs: Synthesis reveals controls over cross-ecosystem 10.1038/nature12760 resource use. Sci. Adv. 3: e1601765. doi:10.1126/ Roberts, B. J., and P. J. Mulholland. 2007. In-stream biotic sciadv.1601765 control on nutrient biogeochemistry in a forested stream, Tank, J. L., and W. K. Dodds. 2003. Nutrient limitation of West Fork of Walker Branch. J. Geophys. Res. Biogeosci. epilithic and epixylic biofilms in ten North American 112: G04002. doi:10.1029/2007JG000422 streams. Freshw. Biol. 48: 1031–1049. doi:10.1046/j.1365- Roberts, B. J., P. J. Mulholland, and W. R. Hill. 2007. Multi- 2427.2003.01067.x ple scales of temporal variability in ecosystem metabolism Uehlinger, U. 2000. Resistance and resilience of ecosystem rates: Results from 2 years of continuous monitoring in a metabolism in a flood-prone river system. Freshw. Biol. forested headwater stream. Ecosystems 10: 588–606. doi: 45: 319–332. doi:10.1111/j.1365-2427.2000.00620.x 10.1007/s10021-007-9059-2 Uehlinger, U. 2006. Annual cycle and inter-annual variabil- Rode, M., S. H. N. Angelstein, M. R. Anis, D. Borchardt, and ity of gross primary production and ecosystem respiration M. Weitere. 2016. Continuous in-stream assimilatory in a floodprone river during a 15-year period. Freshw. nitrate uptake from high frequency sensor measurements. Biol. 51: 938–950. doi:10.1111/j.1365-2427.2006.01551.x Environ. Sci. Technol. 50: 5685–5694. doi:10.1021/ Uehlinger, U., and M. W. Naegeli. 1998. Ecosystem metabo- acs.est.6b00943 lism, disturbance, and stability in a prealpine gravel bed

S117 Bernhardt et al. Metabolic regimes

river. J. North Am. Benthol. Soc. 17: 165–178. doi: stream syndrome: Current knowledge and the search for a 10.2307/1467960 cure. J. North Am. Benthol. Soc. 24: 706–723. doi: Uehlinger, U., M. Naegeli, and S. G. Fisher. 2002. A hetero- 10.1899/04-028.1 trophic desert stream? The role of sediment stability. Webster, J. R., and J. L. Meyer. 1997. Organic matter budgets West. N. Am. Nat. 62: 466–473. for streams: A synthesis. J. North Am. Benthol. Soc. 16: Ulseth, A. J., E. Bertuzzo, G. A. Singer, J. Schelker, and T. J. 141–161. doi:10.2307/1468247 Battin. 2017. Climate-induced changes in spring snow- Whittaker, R. H. 1962. Classification of natural communities. melt impact ecosystem metabolism and carbon fluxes in Bot. Rev. 28: 1–239. doi:10.1007/BF02860872 an alpine stream network. Ecosystems 23: 1–18. doi: Wootton, J. T., M. S. Parker, and M. E. Power. 1996. Effects 10.1007/s10021-017-0155-7 of disturbance on river food webs. Science 273: 1558– Vannote, R. L., G. W. Minshall, K. W. Cummins, J. R. Sedell, 1561. doi:10.1126/science.273.5281.1558 and C. E. Cushing. 1980. The river continuum concept. Young, R. G., and A. D. Huryn. 1996. Interannual variation Can. J. Fish. Aquat. Sci. 37: 130–137. doi:10.1139/f80-017 in discharge controls ecosystem metabolism along a grass- Vorosmarty, C. J., P. Green, J. Salisbury, and R. B. Lammers. land river continuum. Can. J. Fish. Aquat. Sci. 53: 2199– 2000. Global water resources: Vulnerability from climate 2211. doi:10.1139/f96-186 change acid population growth. Science 289: 284–288. Zarfl, C., A. E. Lumsdon, J. Berlekamp, L. Tydecks, and K. doi:10.1126/science.289.5477.284 Tockner. 2014. A global boom in hydropower dam con- Vorosmarty, C. J., and D. Sahagian. 2000. Anthropogenic struction. Aquat. Sci. 77: 161–170. doi:10.1007/s00027- disturbance of the terrestrial water cycle. Bioscience 50: 014-0377-0 753–765. doi:10.1641/0006-3568(2000)050[0753:ADOTTW] 2.0.CO;2] Acknowledgments Vor€ osmarty,€ C. J., and others. 2010. Global threats to human We wish to thank fellow river metabolism enthusiasts Tom Battin, water security and river biodiversity. Nature 467: 555– Joanna Blaszczak, Natalie Griffiths, Ashley Helton, Brian McGlynn, Eliza- beth Sudduth, Phil Savoy, and Amber Ulseth for the many productive 561. doi:10.1038/nature09440 conversations on these topics that have helped refine the ideas pre- Wallace, J. B., M. R. Whiles, S. Eggert, T. F. Cuffney, G. H. sented here. The ideas discussed here have resulted from collaborations Lugthart, and K. Chung. 1995. Long-term dynamics of among the authors that have been enabled by working group funds coarse particulate organic matter in 3 Appalachian moun- through the USGS Powell Center Synthesis Center and the NSF Macro- tain streams. J. North Am. Benthol. Soc. 14: 217–232. doi: systems Program (Grant # EF 1442439). Individual participants have been supported by postdoctoral funding from the Basque Government 10.2307/1467775 (to M. Arroita) and a Humboldt Foundation Sabbatical Fellowship (to E. Wallace, J. B., S. L. Eggert, J. L. Meyer, and J. R. Webster. Bernhardt). 1997. Multiple trophic levels of a forest stream linked to terrestrial litter inputs. Science 277: 102–104. doi: Conflict of Interest None declared. 10.1126/science.277.5322.102 Wallace, J. B., S. L. Eggert, J. L. Meyer, and J. R. Webster. 1999. Effects of resource limitation on a detrital-based Submitted 18 May 2017 ecosystem. Ecol. Monogr. 69: 409–442. doi:10.1890/0012- Revised 29 August 2017 9615(1999)069[0409:EORLOA]2.0.CO;2] Accepted 14 September 2017 Walsh, C. J., A. H. Roy, J. W. Feminella, P. D. Cottingham, P. M. Groffman, and R. P. Morgan. 2005. The urban Associate editor: Ryan Sponseller

S118